“Zhiyuan Tang”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
第11行: 第11行:
 
1. transfer learning(dark knowledge) for nnet3;
 
1. transfer learning(dark knowledge) for nnet3;
  
2. speech & speaker recognition joint training.
+
2. joint training for speech & speaker recognition(ASSR) .

2016年2月22日 (一) 00:24的版本

Last week:

1. With Mengyuan, conducted experiments for small model with big data, initial results here;

2. got the conlusion that mpe improves chain, results here.


This week:

1. transfer learning(dark knowledge) for nnet3;

2. joint training for speech & speaker recognition(ASSR) .